Added value of the tumor-stroma ratio to the PREDICT model for breast cancer

The tumor-stroma ratio (TSR) has repeatedly proven to be a strong prognostic and predictive marker in breast cancer. However, clinical integration of the TSR to further improve patient selection for breast cancer treatment and clinical decision making remains to be defined. Therefore, we evaluated the added value of the TSR to the well-recognized PREDICT model for survival outcomes following adjuvant systemic therapy.

In total, 1770 patients from the Nottingham Breast Cancer Series were included. The incremental value of the TSR following addition to the PREDICT model regarding change in overall prediction model performance was evaluated based on Cox regression analyses and the area under the curve (AUC) for 5-, 10- and 15-year overall survival. Analyses were reported for the total patient group and stratified by molecular subgroups.

The largest model performance improvement was observed in the triple-negative breast cancer (TNBC) subgroup. At 5 years, the AUC ameliorated from 63.9 (95% confidence interval 55.1-72.7) to 70.9 (63.4-78.3) (p = 0.05) with a hazard ratio (HR) of 2.58 (1.39-4.79, p = 0.003). For 10-year survival, the HR was 2.02 (1.21-3.39, p = 0.007), with a rise in the AUC from 64.0 (55.6-72.3) to 69.0 (61.3-76.6) (p = 0.1). At 15 years after clinical diagnosis, the AUC increased from 64.6 (54.0-75.2) to 71.6 (61.1-82.1) at borderline significance (p = 0.06), together with an HR of 2.02 (1.24-3.27, p = 0.005).

The AUC improved for the total patient cohort and all subgroups following the addition of the TSR, although the strongest and most significant model enhancements were observed in the TNBC subgroup.